Surrogate-assisted optimization for decision-making in lithium-ion battery manufacturing

In modern manufacturing systems, digital twins (DTs) have emerged as transformative enablers of smart, adaptive, and data-rich production. A digital twin continuously synchronizes real-time data from physical assets (e.g., machines, sensors, and operators) with a virtual counterpart that mirrors the dynamic behavior of the production system. Such capabilities are particularly crucial in lithium-ion battery production, a process characterized by multi-stage workflows, nonlinear process dependencies, and tight interrelations between quality, energy efficiency, and time. Within this context, digital twins provide a virtual environment for testing parameter adjustments, anticipating defects, and minimizing downtime and contributing directly to the Zero-Defect Manufacturing (ZDM) paradigm. This vision is also central to the European BATTwin project, which develops a multilevel digital twin platform to enhance sustainability and defect reduction in Li-ion battery gigafactories across Europe. The proposed surrogate-assisted optimization framework demonstrates how lightweight predictive models can significantly accelerate decision-making when embedded within a digital twin for Li-ion battery production. By combining high-fidelity simulation with fast surrogate prediction, the system enables adaptive and robust decision-making while reducing computational cost. Future work will focus on enabling online updates of the surrogate model, incorporating uncertainty quantification, and expanding the optimization to fully multi-objective and real-time settings.